Mixture of Experts

What defined AI in 2025 and what’s coming in 2026? In this special year-end episode of Mixture of Experts, host Tim Hwang sits down with veteran panelists to review the biggest AI moments of 2025 and make bold predictions for the year ahead. Kaoutar El Maghraoui unpacks the AI hardware supply crisis and NVIDIA's chip dominance. Gabe Goodhart defends open source's breakout year with models like Kimi K2 Thinking. Chris Hay argues 2025 delivered on "super agents" through ChatGPT Deep Research and Claude Code. Finally, Aaron Baughman and Abraham Daniels predict the future of multimodal AI—from vision models to autonomous digital workers. Join our experts as they separate AI hype from reality and forecast what 2026 holds. 00:00 – Introduction 00:45 – Chris: Super agents, reasoning models and 2025 reflections 12:26 – Gabe: Open source's breakthrough year and what's next 22:17 – Kaoutar: AI hardware in 2025 and predictions for 2026 29:18 – Aaron & Abe: Multimodal AI and modular model orchestration The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity. Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120 Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts

What’s the future of AI code generation? This week on Mixture of Experts, host Tim Hwang is joined by Chris Hay, Olivia Buzek and Gabe Goodhart to debrief the biggest AI use-case of 2025: AI-powered software engineering. Claude Opus 4.5 solved a months-long optimization in under an hour but failed spectacularly at simple tasks. The barbell effect is real. Next, who's the architect—you or the model? We discuss agent orchestration, context windows and why tool performance varies wildly. Then, model differentiation: are OpenAI and Anthropic fundamentally different, or does agent architecture matter more? Finally, can open-source compete with closed ecosystems? We explore vertical integration, inference costs and the future of open models. All that and more on this week's Mixture of Experts. 00:00 – Introduction 01:11 – The barbell problem: AI coding wins and fails 03:46 – Claude Code cracks Apple Metal optimization 07:52 – Who's the architect: You or the AI? 11:44 – Model vs agent orchestration 20:44 – The future of unsupervised AI agents 24:30 – Open source vs proprietary tools 33:22 – The inference cost challenge The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity. Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120 Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts Learn more about AI code generation → https://www.ibm.com/think/topics/ai-code-generation

Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts Why did Disney pay OpenAI a billion dollars to use their characters? This week on Mixture of Experts, host Tim Hwang and experts Marina Danilevsky, Martin Keen and Kush Varshney analyze Disney's three-year OpenAI licensing deal, and what it means for IP owners, content creators and the future of fan-generated AI content. Next, Time Magazine names “Architects of AI” as 2025 Person of the Year—it’s not the first time the person of the year was not a person, but what’s different about this? Then, NVIDIA drops Nemotron 3 open-source models; we explore what makes this model release different. Finally, Anthropic’s Soul Document leaked. We unpack model alignment, philosophy in AI and the future of prompting vs. fine-tuning. 00:37 – Introduction 02:14 – Disney and OpenAI billion-dollar deal 10:35 – Time Magazine's Person of the Year: Architects of AI 15:39 – NVIDIA Nemotron 3 open-source models 24:10 – Claude's Soul Document and model alignment The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity. Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120 #OpenAI, #DisneyAI, #NVIDIANemotron, #ClaudeAI, #AIModels

Should we care about GPT-5.2? This week on Mixture of Experts, we analyze the “code red” release of GPT-5.2 as OpenAI responds to Gemini 3. Are the constant model drops benefitting consumers? Next, Stanford released their Foundation Model Transparency Index, revealing a troubling trend that most labs are becoming less transparent. However, IBM Granite achieved a 95/100 score. Then, our experts discuss what model transparency means for enterprise AI adoption. Finally, we debrief AWS re:Invent’s biggest announcements, including Nova frontier models and Nova Forge. Join host Tim Hwang and panelists Kate Soule, Ambhi Ganesan and Mihai Criveti for our expert insights.00:00 – Intro1:02 -- GPT-5.2 emergency release 12:21 -- Stanford AI Transparency Index: Granite scores 95/10027:18 -- AWS re:Invent: Nova models and enterprise AIThe opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity.Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts #GPT-5.2 #AITransparency #GraniteModels #AWSNova #AIAgents

Is open source winning the AI race? This week on Mixture of Experts, we analyze three major model releases that dropped in the final weeks of 2025: Mistral 3, DeepSeek-V3.2 and Claude Opus 4.5. Our experts discuss what makes each model unique—from Mistral’s multimodal capabilities to DeepSeek’s reasoning-first approach and Claude’s developer focus. Are there too many good models? Next, a provocative blog post from Theory Ventures argues Gemini 3 proves scaling laws are throwing more compute at the problem. We debate if scaling laws are a universal truth. Finally, Amazon just blocked ChatGPT’s shopping research agent from accessing product data. We discuss the business incentives threatening the agent dream. Join host Tim Hwang and panelists Aaron Baughman, Abraham Daniels and Gabe Goodhart on this week’s Mixture of Experts for more! 00:00 – Intro 02:05 -- Model launches: Mistral 3, DeepSeek-V3.2 and Claude Opus 4.5 15:32 -- AI scaling laws & Gemini 3 26:23 --Amazon blocking ChatGPT shopping research agent The opinions expressed in this podcast are solely those of the participants and do not necessarily reflect the views of IBM or any other organization or entity. Visit Mixture of Experts podcast page to get more AI content → https://www.ibm.com/think/podcasts/mixture-of-experts Subscribe for AI updates → https://www.ibm.com/account/reg/us-en/signup?formid=news-urx-52120 #Mistral3 #DeepSeek #ClaudeOpus #AIscalinglaws #AIagents